• Dev Notes
  • Posts
  • The OSI's New "Open Source AI" Definition Raises Eyebrows

The OSI's New "Open Source AI" Definition Raises Eyebrows

PLUS: 2024's Programming Language Landscape-JavaScript Still Reigns Supreme

Good Morning! The Open Source Initiative is stirring the pot with their new "Open Source AI" definition that takes a surprisingly flexible stance on training data transparency. In the ever-evolving world of programming languages, JavaScript continues its reign supreme in 2024, commanding 31% of developer job postings while Go shows some unexpected upward momentum. GitHub Copilot is getting a major upgrade with the addition of Claude 3.5 and Gemini models to its toolbox, letting developers mix and match AI assistants for different coding tasks.

The OSI's New "Open Source AI" Definition Raises Eyebrows

The Open Source Initiative (OSI) just dropped their Release Candidate 1 for the "Open Source AI Definition" – and it's stirring up quite a debate in the tech community. The OSI, known for being the gatekeeper of what qualifies as "open source" in software, is now trying to tackle the AI space.

The definition promises four key freedoms for AI systems:

  • Use without permission

  • Study system components

  • Modify for any purpose

  • Share freely with or without modifications

The Controversy: Here's where things get spicy – the definition's handling of training data is raising red flags. Unlike traditional open source requirements that demand complete source code access, the AI definition only asks for "sufficiently detailed information" about training data. It even allows for "unshareable" data to be used, as long as it's documented.

Tech Take: This looser standard effectively creates a loophole that could legitimize AI systems trained on potentially problematic or illegally scraped datasets. For large-scale AI models, true open source might be practically impossible – they're typically trained on massive datasets that can't be fully shared or legally redistributed. The new definition feels less like traditional open source and more like "AI freeware" – you get the weights and basic docs, but not the full picture needed for true reproducibility.

Read More Here

2024's Programming Language Landscape-JavaScript Still Reigns Supreme

The latest analysis from DevJobsScanner, covering 12M developer jobs from January 2023 to September 2024, reveals some fascinating shifts in programming language demand. The study specifically tracked job listings that explicitly mentioned programming languages in their titles, providing a clear picture of market preferences.

What's Hot: JavaScript/TypeScript continues to dominate the dev landscape, claiming a whopping 31% of language-specific job postings (651K jobs). This shouldn't surprise anyone following the evolution of web technologies, especially with TypeScript's growing adoption in major frameworks like Angular and NestJS. Python secured second place with 20% (408K jobs), showing particular strength in areas like data analysis and server-side development.

While Java maintains its position in the top three with 18% market share, it's worth noting the growing gap between Java and Python. C# holds steady at 12%, powered by its .NET ecosystem and Unity game development. Perhaps the most intriguing shift is C/C++'s decline – dropping from 10% to around 6-7% market share since early 2023. Meanwhile, Go is showing unexpected upward momentum, with a notable spike in August 2024 breaking its usual 2% pattern.

For developers planning their next career move, these trends suggest that investing time in JavaScript/TypeScript and Python remains a solid strategy, while keeping an eye on emerging opportunities in Go could prove valuable.

Read More Here

GitHub Copilot Embraces Multi-Model AI: Adding Claude 3.5 and Gemini to the Mix

GitHub is shaking things up by introducing a multi-model approach to Copilot. Starting with Copilot Chat (web and VS Code), they're rolling out Anthropic's Claude 3.5 Sonnet in the coming weeks, followed by Google's Gemini 1.5 Pro. They're also expanding their OpenAI model options to include GPT o1-preview and o1-mini, which supposedly pack more reasoning punch than GPT-4.

Here's what makes this update particularly interesting for developers:

  • You can switch between models mid-conversation, letting you leverage each model's strengths for different tasks (think: using one model for Python and another for Rust)

  • Organizations can set model accessibility permissions for their teams

  • Multi-model support is coming to more Copilot features, including multi-file editing, code review, and security autofixes

Bonus Round: GitHub's also teasing "Spark," a natural language app development tool that could make coding more accessible to non-programmers while giving experienced devs another tool in their arsenal. It's in early preview now, with a waitlist for those interested in taking it for a spin.

Read More Here

🔥 More Notes

📹 Youtube Spotlight

Coding a Realtime App with TypeScript

Forrest Knight

Was this forwarded to you? Sign Up Here

Reply

or to participate.